OWASP LLM
LLM06 · Excessive Agency
Agents hold more functionality, permissions or autonomy than needed, so a manipulated model can do real damage.
OWASP Top 10 for LLM Applications (2025)
Mitigations
- Least-privilege tools and credentials
- Human approval for high-impact actions
- Rate limits and audit logs for tool calls
Detections 4
Related techniques
- ASI02 Tool Misuse and Exploitation OWASP Agentic
- ASI03 Identity and Privilege Abuse OWASP Agentic
In the news
-
Intent Contracts: Taking Tools Away from Agents
Researchers introduce intent contracts, an authorization interface that removes tools from the planning model in tool-using LLM agents and routes calls through a trusted model.
ReviewReview tool authorization for LLM agents against planner-driven command risks
Tools
Test or exercise it
- promptfoo LLM red-teaming · Adopt · Dual-use
Detect or mitigate it
- NeMo Guardrails Guardrail · Adopt
- promptfoo LLM red-teaming · Adopt
Mappings are AI-assisted and checked against the framework definitions; corrections welcome.